how to use 4mc/4mz when write to json or parquet

I am able to load json files compressed by 4mz/c into spark. But writing does not work. Is this expected?

```
In [53]: df.write.mode('overwrite').option("codec","com.hadoop.compression.fourmc.ZstdCodec").csv('foo')

IllegalArgumentException: 'Codec [com.hadoop.compression.fourmc.ZstdCodec] is not available. Known codecs are bzip2, deflate, uncompressed, lz4, gzip, snappy, none.'
```

